The set of main verbs can be characterized semantically by the fact that they function as n-place predicates denoting certain states of affairs; cf. Section 1.2.3 for a more detailed discussion of the latter term, which is a cover term for states and several types of events.

This semantic property is reflected syntactically by the fact that main verbs normally function as argument-taking heads of clauses. That main verbs function as heads of their clauses is clear from the fact that they are usually indispensable; the primeless examples in (13) would not normally be recognizable as clauses without the verb. The arguments of the verbs are of course needed to express a proposition, but they are not as indispensable as the verb, as will be clear from the fact that the imperatives in the primed examples are perfectly acceptable despite the fact that the arguments of the verb remain implicit.

That main verbs function as semantic heads of clauses is also clear from the fact that clauses contain at most a single main verb; sentences with more than one main verb are usually construed as containing more than one clause. For instance, the primed examples in (14) are cases of embedding: the bracketed clauses can be analyzed as embedded direct object clauses of the matrix verbs vermoedento suspect and vertellento tell. This analysis is supported by the fact that the embedded clauses can be pronominalized, as shown in the primed examples.

Since copular verbs can occur as the only verb of a clause, they are usually also considered main verbs, even though they do not fulfill the semantic criterion of denoting states of affairs; they are not n-place predicates on a par with the predicates in (12), but rather resemble the non-main verbs discussed in the next subsection in that they express specific temporal, aspectual, or modal notions. The copular verb zijn in (15a), for example, locates the state expressed by the proposition ill(Jan) at a certain position on the time axis by carrying a tense marking [±past]: the present tense marking on the finite verb is in (15a) expresses that the state of Jan being ill is true at speech time. The copulas worden and blijven in (15b) express additional aspectual information: wordento become is mutative in that it indicates that Jan is in the process of acquiring the state of being ill; blijvento stay is in a sense the opposite of worden in that it expresses that the state of Jan being ill continues to exist. Copular verbs like lijkento appear or blijkento turn out in (15c) are modal in nature, since they specify the speaker’s attitude towards the truth value of the proposition.

Although the distinction between main and non-main verbs seems relatively clear, it is not easy to give an operational definition of this distinction, so it is not surprising that grammars of Dutch may differ in where they draw the line between the two categories. Like many other Dutch grammars, Haeseryn et al. (1997:46) assumes that main verbs are predicative, i.e. “express the core meaning of the verbal complex”, while non-main verbs function as modifiers that provide additional information; they give the verb types in (16) as typical examples of non-main verbs. In order to fully appreciate what follows, it is necessary to point out that we used the term verbal complex in the quote above as a translation of the Dutch term werkwoordelijk gezegde from traditional grammar, which is not easily translated into English.

Haeseryn et al. (1997:47) tries to use the essentially semantic characterization of main and non-main verbs in order to provide an operational definition in syntactic terms. The crucial criterion they mention is that each verbal complex contains exactly one main verb. If we apply this criterion to a perfect-tense or passive example, this characterization goes two ways: if we assume that the participles in (17) are main verbs, we should conclude that the temporal/passive auxiliaries are non-main verbs; if we assume that the temporal/passive auxiliaries are non-main verbs, we should conclude that the participles are main verbs.

The one-main-verb criterion implies that main verbs differ crucially from non-main verbs in that they can, but do not have to, be combined with other verbs to form a verbal complex, whereas non-main verbs must always be combined with some other verb. This seems to work well in the case of the examples in (17): the examples in (18) show that the verb aaiento pet can indeed occur as the only verbal element in a clause, whereas the temporal and passive auxiliaries cannot (although hebben can be used as a main verb meaning “to have/possess” and worden as a copular verb meaning “to become”, hence the use of the number sign).

The one-main-verb criterion also fits well with our intuition that in examples such as (19) we are dealing with two predicational relations and thus with two verbal complexes. That the verb horento hear functions as a separate predicate can be seen from the pronominalization of the italicized phrase in (19a); since horen is the only verb in (19b), it must be a main verb.

However, if we apply the pronominalization test to the examples in (20), we have to conclude that modal verbs like moetenmust and mogenmay are also main verbs. This means that we can only maintain the earlier claim that modal verbs are non-main verbs if we are willing to assume that clauses with modal verbs are exceptions to the general rule that non-main verbs must be combined with a main verb; cf. Klooster (2001:55) for discussion.

There are many complications with the claim that modal verbs are non-main verbs. For example, it implies that we should conclude that example (19a) contains two separate verbal complexes, whereas example (20a) has only one verbal complex, but there is no independent syntactic evidence to support this. On the contrary, the obligatory clustering of the finite and non-finite verbs of (19a) and (20a) in the embedded clauses in (21) shows that both cases involve a single verbal complex. In fact, this would imply that the one-main-verb criterion would incorrectly characterize the verb horento hear in (19a) as a non-main verb.

For English it could perhaps be argued that modals such as can are non-main verbs because, as shown in (22), they are like perfect auxiliaries in that they can precede negation and can be inverted with the subject in e.g. questions (although this may also be a side effect of the accidental morphological property of modal verbs that they have no infinitival form, as is clear from *to can, which makes them incompatible with do-support). See Quirk et al. (1979:120ff) and Huddleston & Pullum (2002:92ff) for an overview of the criteria used to identify English auxiliaries, and Palmer (2001:100) for a more specific discussion of English modal verbs. In Dutch, there is no such syntactic evidence that the modal verbs in (16c) are different from the perfect and aspectual verbs in (16a&c).

We will not follow the characterization of the distinction between main and non-main verbs based on the one-main-verb criterion, since the discussion above has shown that this criterion is highly problematic and empirically inadequate. So, we simply assume that any verb must be considered a main verb that is predicative (i.e. has an argument structure) and thus can function as the head of its own (finite or infinitival) clause. This in turn implies that verbal complexes can contain more than one main verb and it reduces the set of non-main verbs by excluding modal verbs such as moetenmust. We refer the reader to Section 4.6 for a more detailed discussion of the distinction between main and non-main verbs.